14 nm ARM chip taped out

[Fjodor2001]

See:

http://www.marketwatch.com/story/ca...-and-ibm-finfet-process-technology-2012-10-30

 

[MisterMac]

Forgot to mention it's the M0 Design.

Hardly the most complex IC to fab.


Maybe IDC can shed on wether this is significant comparing to Intel?

 

[ShintaiDK]

Its a test chip. And with IBM as well it looks to be nothing more than the usual IBM PR stunt.

How many transistors is the M0 again? ARM itself writes under 12K gates. At 40nm its below 0.01mm2 in size.

 

[Idontcare]

Maybe IDC can shed on wether this is significant comparing to Intel?

It is a milestone that must be checked off in the course of 14nm-XM reaching production. Other than that, nothing really interesting going on there.

Now if they had come out with some actual first-silicon results to speak to then that would have been noteworthy.

This is GloFo though, and if you went by their past PR then at best this means they are a solid 2yrs away from putting 14nm-XM into production.

Jan 9, 2010 by Charlie Demerjian

GLOBAL FOUNDRIES has been way ahead of the curve with process tech when compared to any other foundry on the market. Today, it somewhat accidentally cemented that lead by publicly displaying an unlabeled 28nm wafer.

28nm wafers that are not test structures

Global Foundries representatives would not talk about what chips were on that wafer, but they were definitely not the test SRAM structures that we saw in June. The new wafers were quite irregular in appearance, so they could be CPU or GPU chips, or they could be some very advanced test structures. Given the size ballparks at over 300mm^2, they are unlikely to be ARM cores, but they could be advanced SOCs based on ARM designs.

In any case, Global Foundries 28nm process is looking good. Officially, it is still on track for the second half of 2010, but given this showing, that may be an overly cautious timetable. Progress like this tends to lead to a slew of new products, so bring on the new toys.S|A

http://semiaccurate.com/2010/01/09/global-foundries-28nm-wafer-spotted/

^ we are nearly 3yrs past this S/A article on how teh awesome GloFo was and how bitchen their 28nm was coming along

So what are we expected to take away from an article on a tapeout of a device that could someday be fabbed on GloFo's 14nm-XM? I'll believe when I see the results

 

[ShintaiDK]

Oh good find IDC. LOL!

Whenever I read some PR about IBM and GLoFo, you always know its a heavily modified truth.

Another favourite of mine was the 250Ghz transistors. Just a shame it was only a handful.

 

[NostaSeronx]

GlobalFoundries had successful tape outs of Krishna and Wichita in 2011. GloFo also had successful tape outs of Kaveri and Richland in 2012. Most of these parts were cancelled for the same reason why Northern Islands wasn't on GlobalFoundries 32-nm. GlobalFoundries was and still is charging an arm and a leg for services because they haven't been profitable.

 

[Idontcare]

GlobalFoundries had successful tape outs of Krishna and Wichita in 2011. GloFo also had successful tape outs of Kaveri and Richland in 2012. Most of these parts were cancelled for the same reason why Northern Islands wasn't on GlobalFoundries 32-nm. GlobalFoundries was and still is charging an arm and a leg for services because they haven't been profitable.

I don't follow this, globalfoundries being the foundry doesn't do tapeouts. Their customers do tapeouts. Regardless, AMD did not tapeout Krishna or Wichita, they were cancelled pre-tapeout. Likewise Kaveri has not taped out.

Richland supposedly has taped out, but I underscore the supposedly there because at most that is a strong rumor which is questionable at best because AMD pretty much always makes a big deal out of tapeouts for major products to ensure the analysts know, but they have not announced tapeout for Richland yet (or Kaveri, or Jaguar-based SoCs for that matter).

 

[phenomkid7]

Richland supposedly has taped out, but I underscore the supposedly there because at most that is a strong rumor which is questionable at best because AMD pretty much always makes a big deal out of tapeouts for major products to ensure the analysts know, but they have not announced tapeout for Richland yet (or Kaveri, or Jaguar-based SoCs for that matter).

AMD has to launch Kabini in Q3 2013, or go out of business. So what do you mean they haven't taped out Kabini? Unless I misunderstand what this archaic term means...but I thought the time between tapeout and manufacture is 12-14 months.

 

[Idontcare]

AMD has to launch Kabini in Q3 2013, or go out of business. So what do you mean they haven't taped out Kabini? Unless I misunderstand what this archaic term means...but I thought the time between tapeout and manufacture is 12-14 months.

If it has taped out, and that is the operating assumption that I am willing to go with, the observation to be made here is that AMD has elected to be atypically quiet about said taping out.

Tapeout events are big deals to mention to analysts despite the relatively little big-picture relevance of hitting the milestone itself (getting through the verification stage is the biggie) and yet AMD has not mentioned Kabini being taped out TMK.

So we can take this to mean it hasn't happened yet (Occam's razor says this is the case), or we can begin concocting schemes in our minds to justify AMD electing to keep it a secret all along.

The last time I found out AMD was electing to keep a secret it turned out the secret was that IPC really did decrease.

 

[NostaSeronx]

I don't follow this, globalfoundries being the foundry doesn't do tapeouts. Their customers do tapeouts. Regardless, AMD did not tapeout Krishna or Wichita, they were cancelled pre-tapeout. Likewise Kaveri has not taped out.

Krishna and Wichita were cancelled post-tapeout and Kaveri as well was cancelled post-tapeout. The only 28-nm APU to sucessfully tape out and go into production at GlobalFoundries is Richland and only Richland.

 

[Subyman]

ARM never taps out bro

 

[Phynaz]

Krishna and Wichita were cancelled post-tapeout and Kaveri as well was cancelled post-tapeout.

Source?

I'm curious to see if you provide one this time, as the last three times I asked for one you conveniently disappeared.

 

[CTho9305]

I don't follow this, globalfoundries being the foundry doesn't do tapeouts. Their customers do tapeouts. Regardless, AMD did not tapeout Krishna or Wichita, they were cancelled pre-tapeout. Likewise Kaveri has not taped out.

Richland supposedly has taped out, but I underscore the supposedly there because at most that is a strong rumor which is questionable at best because AMD pretty much always makes a big deal out of tapeouts for major products to ensure the analysts know, but they have not announced tapeout for Richland yet (or Kaveri, or Jaguar-based SoCs for that matter).

Xbitlabs says they had commented already.

 

[Idontcare]

Xbitlabs says they had commented already.

I missed that, thanks for the link :thumbsup: